• Bitwise matrix multiplication
• Computational efficiency
• cppyy
• Matrix Calculation
• Neural network optimization
• SIMD instructions
• tests

This thesis focuses on the innovative domain of matrix multiplication optimizations through binarization techniques, starting from the seminal work of Nardini et al.(“Neural Network Compression using Binarization and Few Full-Precision Weights”). At the heart of this thesis lies the quest to translate and adapt the C++ methodologies described in the article into Python.
The process of transposing the binary matrix multiplication function and its auxiliary components - 'pack into 64 bits' and 'reorganize' - into Python was not merely a technical exercise, but a conceptual redesign.
Through the use of cppyy, a library facilitating C++ and Python interoperability, we achieved a seamless integration that preserves the computational efficiency of the original C++ code while leveraging Python’s accessibility and flexibility.
Our empirical investigations highlight significant differences between C++ and Python implementations in terms of execution time.
In conclusion, this thesis not only contributes to the academic discourse through its methodological innovations and empirical findings, but also beckons future research to explore uncharted territories in computational efficiency and AI application. The integration of binarized matrix multiplication into the broader landscape of neural network design and optimization stands as a testament to the transformative potential of bridging theoretical insights with practical implementations.